The AI story is splitting in two: speculative smaller names like Rezolve AI are getting sold on any dip, while durable money keeps buying the infrastructure layer — Berkshire now counts an AI hyperscaler among its top three holdings. Oracle has quietly be

Verdict: the AI backbone thesis is intact, but neither trade is armed — wait for the trigger

The idea's core argument — durable money rotating into AI infrastructure — is well supported by the fundamentals: NVDA posted $215.9B revenue, up 65.5% year over year, with $96.7B of free cash flow and debt-to-equity down to 0.05, while ORCL's revenue grew 17.3% to $67.4B with a 30.6% operating margin in the 97th percentile of peers. But the strongest point against the trade as implemented is the completed backtest: over 60 months the fade rules lost 7.7% across 15 trades with a 33.3% win rate and a 10.7% maximum drawdown, and the short-entry rules test something closer to the opposite of the buy-the-pullback thesis. ORCL's negative $23.7B free cash flow (0th percentile among 612 IT peers) and $219.3B of liabilities against $42.5B of equity make it the clear first stress point if data-center returns disappoint. No robust parameter setup was established, since the sensitivity evaluation ran out of its time budget. Today neither entry is live: NVDA's RSI (14) at 42.1 sits 7.9 points below its 50 trigger, and ORCL's ADX at 8.5 sits 16.5 points below its 25 trigger — so the honest action is to wait, with NVDA the closer of the two catalysts.

Trade now

## Trade now: wait for the momentum trigger This is a contrarian fade setup: it shorts NVDA and ORCL when they tap their 50-day EMA from below and reclaim it, with momentum confirmation. **Neither entry is live today.** **NVDA** ($211.7): price is hovering right at the 50-day EMA of $210.75 — just $0.95 above it — and the trend-strength condition is already met, with ADX at 50.8 against a 25 threshold. The missing piece is the RSI (14) cross above 50: it reads **42.1 now**, about 7.9 points below the trigger. That is the binding condition. **ORCL** ($147.26): the price conditions are already met — the close is within $0.85 of the 50-day EMA at $148.11 — and RSI (14) at **56.3** is above 50. The blocker is trend strength: ADX reads **8.5**, a full 16.5 points below the required 25. This tape is drifting, not trending, so the setup stays dormant. **Risk math once triggered:** each position carries an invalidation stop at a 2.6% adverse move and a take-profit at 5.2% — an effective 2:1 reward-to-risk, sized so no position exceeds 25% of capital. A 60-session time stop closes anything that goes nowhere. The completed backtest across both tickers over 60 months (15 trades, 33% win rate) still supports the decision framework as tested. **"Wait" means:** keep alerts on NVDA's RSI crossing 50 while price holds above $210.75, and ORCL's ADX reaching 25 while price holds above $148.11. Do nothing until both legs of the relevant ticker fire on the same signal.

The infrastructure layer has the balance sheets to absorb the pullbacks

The idea's core claim — that durable money is flowing to the AI infrastructure layer rather than the hype names — is visible in the fundamentals. Nvidia's latest fiscal year (ended January 2026) shows revenue of $215.9B, up 65.5% year over year, with a 71.1% gross margin and a 60.4% operating margin that places it in the 99th percentile of its Information Technology peer group. Free cash flow of $96.7B also sits in the 99th percentile. These are the economics of a company that can fund its own growth without leverage: debt-to-equity has fallen steadily from roughly 0.44 in fiscal 2023 to just 0.05 at fiscal year-end 2026. Oracle's transformation matches the thesis that it has quietly become an AI infrastructure play, per the Yahoo Finance piece titled "Oracle Stock Is Now An AI Infrastructure Play." Revenue reached $67.4B in the fiscal year ended May 2026, up 17.3% year over year — its fastest top-line pace in over a decade — with a 30.6% operating margin in the 97th percentile of peers and a 40.2% return on equity. Long-duration data-center contracts are the mechanism the idea cites, and the revenue acceleration is consistent with that story. The competitive-landscape point also holds: the Cerebras-versus-Nvidia coverage (per Yahoo Finance) argues the battle…

Scores

• Conviction score breakdown: 49
• Thesis support: 70
• Trade readiness: 30
• Risk quality: 55
• Backtest evidence: 25
• Fundamentals trend: 65

Watch items

• NVDA — RSI (14)
• ORCL — ADX (14)
• NVDA — Price vs EMA (50)
• ORCL — Price vs EMA (50)
• ORCL — Nearest support
• NVDA — Price
• NVDA — Price above EMA (50)
• NVDA — RSI (14) crossed above 50
• NVDA — ADX (14) above 25